[Intel-gfx] Critical regression in 4.7-rcX

[Larry Finger]

On 07/15/2016 11:36 AM, Bish, Jim wrote:
> On Fri, 2016-07-15 at 10:34 -0500, Larry Finger wrote:
>> On 07/15/2016 09:43 AM, Ville Syrjälä wrote:
>>>
>>> On Fri, Jul 15, 2016 at 04:27:27PM +0200, Daniel Vetter wrote:
>>>>
>>>> On Fri, Jul 15, 2016 at 8:55 AM, Linus Torvalds
>>>> <torvalds at linux-foundation.org> wrote:
>>>>>
>>>>> On Thu, 14 Jul 2016, Larry Finger wrote:
>>>>>>
>>>>>> To anyone keeping track of regressions in kernel 4.7, I call
>>>>>> your attention to
>>>>>> https://bugs.freedesktop.org/show_bug.cgi?id=96675.
>>>>>>
>>>>>> This bug causes driver i915 to fail to connect to the
>>>>>> display, and results in
>>>>>> a blank screen as soon as the kernel is loaded. The only way
>>>>>> to operate with
>>>>>> kernel 4.7 is to add "nomodeset" to the command line. The
>>>>>> problem was bisected
>>>>>> to commit f21a21983ef13a ("drm/i915: Splitting
>>>>>> intel_dp_detect").
>>>>> Daniel? Jani? Time to revert?
>>>> Yup. Means that I'll get to revert a pile of patches (plus even
>>>> more
>>>> in -next), but that's what we get for not handling regressions
>>>> timely.
>>>> Dave poked me about the same bug already too. But will take a few
>>>> days
>>>> until the pull is in your inbox I guess.
>>> I attached a oneliner to the bug today, that I *think* should fix
>>> the
>>> problem.
>> After I checked out the problem commit and built that kernel with the
>> one-line
>> patch, the display works; however, mainline still fails. I am
>> currently
>> bisecting to find what other commit fails. At a minimum, we should be
>> able to
>> reduce the number of commits that need reverting.
>>
>> Larry
> I was seeing the same issue Linus reported but very sporadically.
>  Assumed it was due to my wierd configuration - eDP + usb type c -> MST
> hub -> DVI KVM -> dual head HDMI.  Checked out the one liner - works
> for me but I guess there are still other failures.

I'm not sure what I did wrong before, but when I retested mainline with the one 
liner, it now works. Once that is pushed to mainline, the bug should be fixed.

Sorry for any confusion.

Larry